Antigen selection and cloning

Sixteen sequences from HHV proteins were selected as antigens for HHV species-specific antibody measurement (Table 1). Most immunogenic and species-specific regions were chosen according to the literature. Whenever possible, signal peptides and transmembrane regions were excluded from the recombinantly expressed proteins. For VZV antigens and EBNA-1 peptide, selected inserts (Table 1) were assembled into pGEX4T3tag vector (modified from pGEX4T3) via gene synthesis (eurofins Genomics, Ebersberg, Germany) [27]. These constructs were also codon-optimized for expression in E. coli (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of selected HHV antigens.

https://doi.org/10.1371/journal.pone.0209379.t001

Sequences for all other antigens were derived from genomic DNA (HSV-1, HSV-2, CMV) and cosmid clones (EBV), and used as templates for amplification via PCR. Corresponding primers were designed and subsequently ordered (eurofins Genomics). Sequences for viral antigens were cloned into pGEX4T3tag vector. HHV constructs were amplified in E. coli DH5α, commercially sequenced (eurofins Genomics / GATC Biotech, Konstanz, Germany) and validated against reference sequences from the NCBI nucleotide data base (Table 1).

Antigen expression

Antigens were expressed in E. coli strain BL21 as described previously [27]. Briefly, antigens were expressed as Glutathione S-Transferase (GST) fusion proteins (GST-X-tag) in pGEX4T3tag vectors encoding N-terminally for GST and C-terminally for the last eleven amino acids of SV40 large T-antigen (tag). pGEX4T3tag carries the amp^R gene for positive selection of transformed bacterial colonies in ampicillin-containing medium. In case of VZV, GST-gI-tag, expression was additionally performed in pDB.GST vector (DNASU, Arizona State University Biodesign Institute, DNASU Plasmid Repository, Tempe, Arizona, USA) encoding for a kanamycin resistance instead of ampicillin for selection of co-transformed colonies (co-expression of gE and gI). In both vectors, transcription of GST fusion proteins is inducible by IPTG (tac promotor).

After bacterial cell lysis, lysates were cleared and stored at -20°C in 50% (v/v) glycerol. Quality control of expressed antigens was performed as described previously [22, 27] and included protein gel electrophoresis followed by Coomassie staining, western blot staining for both C- and N-terminal tags to ensure expression of full length antigens and GST capture ELISA for relative quantitation of the fusion proteins [27].

Reference panels and reference assays

Human reference sera and details on the gold-standard reference assays are shown in Table 2. The numeration of the reference panels (RP) corresponds to the HHV numbering in Roman numerals (I-V). For CMV, two reference panels tested with different reference assays were available (denoted as reference panels Va and Vb). The reference sera were obtained from the Institute of Virology and Immunobiology of the University of Würzburg (HSV, EBV, CMV), from the Detroit Neighborhood Health Study (DNHS: HSV-2, CMV) and the Zoster Associated Pain (ZAP) and Shingles UK (SUK) studies (VZV). Sera were sent to the DKFZ on dry ice and were stored at -20°C until testing. The serum collections are described in detail below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of reference serum panels.

https://doi.org/10.1371/journal.pone.0209379.t002

The reference serum panel obtained from the University of Würzburg was composed of two subgroups. Group 1 consisted of serum samples (n = 197; median age 16.9 years (range 0.3–84.6 years); 53% male) received by the viral diagnostic laboratory between 2007 and 2014 for analysis of herpesvirus IgG antibodies. This was part of routine work-up in patients before solid organ transplantation and in patients with malignant diseases before chemotherapy and potentially stem cell transplantation. Group 2 consisted of serum samples (n = 22; median age 1.2 years (range 0.4–25.6 years); 64% male) that were found to be negative for HHV6 IgG antibodies in routine diagnostic testing. With few exceptions, all samples were tested for IgG antibodies against HSV, EBV, and CMV at the viral diagnostic laboratory. The sera were stored at -20°C before shipping to the DKFZ. The use of human serum samples in this study was approved by the ethics committee of the medical faculty at the University of Würzburg. The need for consent was waived by the ethics committee.The DNHS is a longitudinal study of ecologic factors that may influence mental and physical health in an urban setting. DNHS participants are representative of Detroit, Michigan in terms of age, gender, race, income and educational attainment. Participants provided written informed consent for participation and study was approved by the University of Michigan and the University of North Carolina Institutional Review Board (IRB #13–3999) [28]. The reference samples from DNHS were obtained from Wave 1 participants. Frozen serum samples stored at -70°C were shipped on dry ice to the Stanley Neurovirology Laboratory of the John Hopkins University School of Medicine in Baltimore, Maryland to be tested for serum IgG antibodies to CMV and HSV-2 [29]. For each sample, the antibody levels were expressed as ratio of the optical density of a test sample to that of a standard sample assayed in each test run. Individuals were categorized as seronegative if their ratio value was <1.0 and seropositive if ≥1.0.

The VZV reference sera were collected from multiple sources. In the ZAP and SUK study, subjects presenting with acute zoster were followed up for 12 (ZAP) and 6 months (SUK) with serum samples obtained at four time points [30, 31, 32]. Additional sera from VZV positive asymptomatic blood donors were also included in the reference panel [33]. All samples were obtained under the UCLP DNA biobank ethical framework (REC reference: 17/LO/1530). Participants provided written informed consent.

Monoplex and Multiplex Serology

The reference sera were analyzed for antibodies against selected HHV antigens (Table 1) by species-specific Monoplex and Multiplex Serology, as described previously [19]. Briefly, HHV GST-tag fusion proteins were loaded onto glutathione casein-coated fluorescence-labelled polystyrene beads (COOH-beads xMAP Technology Microspheres, Luminex Corp. Austin, Texas, USA) by in situ affinity purification from lysate. Up to 100 bead sets are distinguishable by the Luminex flow cytometer via different ratios of two fluorescent dyes within the polystyrene microspheres. Loading each antigen onto a specific bead set enables simultaneous measurement of antibodies against different antigens within one reaction vessel.

Detection of bound primary antibodies from serum took place with a biotinlyated goat-α-human IgM/IgG/IgA secondary antibody (1:1000, #109-065-064, Jackson Immunoresearch, West Grove, PA, USA) and subsequent incubation with streptavidin-R-phycoerythrin (1:750, PE-Streptavidin Conjugate, MOSS Inc., Pasadena, CA, USA) as reporter dye. Median Fluorescence Intensities (MFI) from at least 100 detected beads per bead set (e.g. antigen) were calculated for each serum. Monoplex Serology was conducted for each HHV species-specific assay in an individual experiment only comprising the species-specific antigens and GST-tag antigen for background subtraction in dilutions 1:100 and 1:1000. Optimal serum dilution was 1:1000 with the exception of VZV (1:100). In addition, performance of HHV Monoplex Serology assays were assessed in multiplex format by combining them with various pathogen-specific Monoplex Serology assays (e.g. human herpes viruses 6–8, human polyomaviruses, human papillomaviruses, human hepatitis B and C viruses) into a Multiplex Serology panel.

Statistical analysis

The reference sera were tested blinded. The Wellcome Centre for Human Genetics functioned as a trusted third party, and combined the DKFZ testing results with the reference data provided beforehand, thus unblinding the analysis. Each antigen-specific serostatus was determined by applying a cut-off to dichotomize the MFI values into seropositive or seronegative. The final cut-off was determined to result in specificity and sensitivity of at least 85% analogous to Receiver Operating Characteristics analysis. This was achieved by gradually raising a working cut-off from a minimum of 30 MFI (dilution 1:1000) or 50 MFI (dilution 1:100) to optimize specificity and sensitivity. The optimum cut-off was determined to favor specificity, unless a further increase of the cut-off resulted in a disproportional loss in sensitivity. Thus, agreement with the reference assay was maximized. When multiple pathogen-specific antigens were included in the assay, seropositivity against the respective pathogen (denoted as overall pathogen seropositivity) was additionally determined by systematic investigations of antigen combinations. Antigen-specific cut-offs were adapted to optimize agreement with the reference assay if necessary as described above. In addition to sensitivity and specificty, Cohen’s kappa (k) statistics to define agreement with the reference assay were calculated and evaluated as follows: 0.01<k<0.20: slight agreement, 0.21<k<0.40: fair agreement, 0.41<k<0.60: moderate agreement, 0.61<k<0.80 substantial agreement and 0.81<k<0.99: almost perfect agreement [35]. Sensitivity, specificity and kappa statistics including 95% Confidence intervals (CI) were calculated using SAS 9.4.

Comparison of the Monoplex and Multiplex Serology performance on the corresponding reference serum panels was conducted by calculating Intraclass Correlation Coefficients (ICCs) using R 3.5.0 package ‘psych’ [36]. ICC(3,1) plus corresponding 95% CI are reported. ICCs were evaluated as follows: 0.01<ICC<0.49: poor reliability, 0.50<ICC<0.74: moderate reliability, 0.75<ICC<0.89: good reliability, 0.90<ICC<1.00: excellent reliability [37].

Antigen development

Based on reported immunogenicity, antigen coverage by the reference assays and sequence homology, 2-4 antigens were selected for development and validation of HHV species-specific Monoplex Serology assays (Table 1). Antigens were expressed as recombinant GST-fusion proteins as described previously [27]. To ensure correct protein sequence, parental plasmids were sequenced. For most antigens, perfect agreement with the reference sequence (NCBI nucleotide database) was confirmed (Table 1). Only HSV-2 antigen mgGunique showed non-silent nucleotide sequence variations compared to strain HG52 resulting in one amino acid change, one single amino acid deletion and one 14 amino acid deletion.

Comparison of HHV species-specific Monoplex Serology assays with reference serostatus

Six reference serum panels (RP I-IV, Va, Vb; Table 2) were analyzed by the corresponding HHV species-specific Monoplex Serology assay. For CMV, two reference panels using different gold-standard assays were available and tested (Va, Vb). Quantitative antibody reactivities (MFI) for each HHV antigen were compared against the corresponding reference serostatus (Fig 1). Where multiple species-specific antigens were included, overall seropositivity in Monoplex Serology was calculated by a combination of the included antigens. Cut-offs were determined by optimizing sensitivity and specificity. The performance characteristics (i.e. specificity, sensitivity and kappa statistics) for the Monoplex Serology assays for HHV 1–5 compared with the gold-standard reference assays are shown for each antigen (Table 3) and overall seropositivity (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Comparison of quantitative antibody measurements (MFI) with reference serostatus in HHV species-specific Monoplex Serology.

I-Vb indicate corresponding reference panels. gE/gI: co-loading of antigens gE and gI onto one bead set red lines: optimized cut-offs for single antigen performance; cut-offs were determined by optimizing specificity and sensitivity. MFI: Median Fluorescence Intensity.

https://doi.org/10.1371/journal.pone.0209379.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Single antigen performance compared to corresponding reference panels in Monoplex Serology.

https://doi.org/10.1371/journal.pone.0209379.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Overall HHV species-specific performance compared to corresponding reference panels in Monoplex Serology.

https://doi.org/10.1371/journal.pone.0209379.t004

HSV species-specific Monoplex Serology validation

Two antigens each were evaluated for their ability to determine HSV species-specific serostatus, HSV-1 gG and gD, and HSV-2 mgGunique (mgGu) and gD. Evolutionary, the HSV glycoproteins gG evolved differently resulting in an approximately 500 aa region unique to HSV-2 [38]. The sequence identity of the regions common to both HSV species is approximately 50% [38]. For HSV-2, gG is cleaved into a secreted (sgG) and a membrane anchored part (mgG) [38] of which the region unique to HSV-2 (mgGu) was expressed and used as antigen in HSV-2 Monoplex Serology. For HSV-1 and -2 gD, sequence identity is approximately 80%. Thus, antibody responses against gD antigens are of high interest for assessing cross-reactivity between both assays.

HSV-1 and -2 Monoplex Serology assays were validated against a HSV species-unspecific (Enzygnost anti-HSV IgG), and a HSV-2 specific (LIAISON HSV-2 IgG) reference assay in RP I and II, respectively. No HSV-1 species-specific reference panel was available. Thus, an indirect approach using HSV-2 Monoplex Serology validation was pursued for HSV-1 Monoplex Serology validation.

The measured antibody reactivities against HSV-1 and -2 antigens based on RP I are shown in Fig 1-I. Both HSV-1 antigens gD and gG discriminate well between reference assay positives and negatives, resulting in only one false-positive and 5 and 3 false-negatives, respectively. The resulting specificity is 98.8% for both HSV-1 antigens, while the sensitivity is 95.9% for gD and 97.6% for gG (Table 3). Based on RP I, HSV-2 gD also showed good capacity to distinguish between reference assay negatives and positives. However, in comparison to HSV-1 gD, a slightly larger overlap in measured antibody reactivities between reference assay negatives and positives was observed between approximately 10 and 100 MFI. HSV-2 mgGu showed very little seroreactivity in RP I, with only six sera showing antibody responses >100 MFI. Thus, no attempt was made to determine meaningful cut-offs for HSV-2 antigens in RP I.

HSV-2 Monoplex Serology was validated against a HSV-2-specific reference assay based on RP II (Fig 1-II). With a cut-off of 180 MFI, HSV-2 mgGu discriminated very well between reference assay positives and negatives resulting in only one false-positive and 4 false-negatives, yielding a specificity and sensitivity of 97.8% and 93.4%, respectively. Kappa statistics showed almost perfect agreement with the reference assay (k = 0.91) (Table 3). For reference assay positives, similar antibody reactivities were observed for HSV-2 gD and mgGu (approx. 200 to 10,000 MFI). However, high HSV-2 gD antibody reactivities were also observed for a substantial number of reference assay negative samples that were also negative for HSV-2 mgGu. Thus, no cut-off for HSV-2 gD was determined in RP II.

HSV-2 Monoplex Serology was successfully validated based on antigen mgGu. Based on the very low prevalence of HSV-2 mgGu, RP I contained very few (<5%) HSV-2 seropositives. Thus, the antibody reactivities measured with HSV-2 gD most likely represent cross-reactivity with HSV-1 gD antibodies. This was confirmed by a high correlation of antibody reactivities between the homologous HSV-1 and -2 gD proteins observed in both RP I and RP II (S1 Fig). Thus, the gD antigens could not be validated for measurement of species-specific HSV antibodies but can be applied for detection of general (species-unspecific) HSV infection. Additionally, based on the small number of HSV-2 seropositives in RP I, this setting allowed us to indirectly validate HSV-1 Monoplex Serology based on antigen gG (RP I) although no HSV-1 species-specific reference panel was available. This is further supported by the low correlation of gG and mgGu (r = 0.13, S2 Fig).

VZV Monoplex Serology validation

The VZV Monoplex Serology antigen panel included glycoproteins E (gE) and I (gI) and immediate early protein 63 (IE63) and was validated against the time resolved fluorescence immunoassay (TRFIA) [31]. Antigen gE differentiated well between reference assay negatives and positives, with only 4 false-positives and 9 false-negatives (Fig 1-III); specificity was 95.2% and sensitivity was 90.7% (Table 3). Kappa statistics showed almost perfect agreement with the reference assay (k = 0.86). gI Monoplex Serology resulted in a substantial number of false-positives and even more false-negatives compared to the reference assay (Fig 1-III). Using a cut-off to yield the minimum desired specificity of 85.0%, calculated specificity was 89.2% and sensitivity was 58.8% (Table 3). Antigen IE63 showed no capacity to discriminate between reference assay negative and positive sera (Fig 1-III). Thus, IE63 is not informative for VZV Monoplex Serology and no cut-off was determined.

VZV proteins gE and gI form hetero-dimers in infected cells [39]. Thus, different approaches were undertaken to assess whether individual antigen performance of gE can be improved by a combination with gI. Determination of overall VZV serostatus by seropositivity to gE and / or gI resulted in very similar statistics compared to gE alone (Tables 3 and 4). To simulate hetero-dimerisation of gE and gI, two additional strategies were pursued to enable detection of VZV antibodies directed against epitopes jointly formed by gE and gI. First, bacterial lysates containing VZV antigens gE and gI were mixed and simultaneously loaded onto the same bead set. Co-loading of gE and gI showed very similar detection characteristics as gE individually (Fig 1-III and Table 3). However, antibody reactivities for both reference assay negatives and positives were slightly increased. In a second approach, antigens gE and gI were co-expressed in E. coli yielding almost identical data compared to co-loading, or gE alone (S3 Fig).

Thus, VZV Monoplex Serology performance is largely driven by antigen gE. In RP III, no added benefit could be achieved by the various approaches to include gI.

EBV Monoplex Serology validation

EBV Monoplex Serology comprises a panel of four EBV proteins. In case of EBV nuclear antigen 1 (EBNA-1), two fragments of differing sizes were expressed, EBNA-1 truncated (EBNA-1 trunc) and EBNA-1 peptide (EBNA-1 pep) (Table 1). In addition, viral capsid antigen p18 (VCA p18), Z-Epstein-Barr virus replication activator (ZEBRA) and early antigen-diffuse (EA-D) were included in the EBV antigen panel. EBV Monoplex Serology was validated against the Enzygnost anti-EBV IgG assay. Among the reference assay seropositive sera (n = 136), 131 (96.3%) were seropositive for EBNA-1 trunc, 124 (91.2%) for VCA p18, 117 (86.0%) for EBNA-1 peptide, 110 (80.9%) for ZEBRA and 103 (75.7%) for EA-D (according to cut-offs shown in Table 4). Antigen-specific concordance with the reference assay is generally good (Fig 1-IV). However, for all antigens between 5 and 22 false-positives and/or false-negatives were observed. Specificity for individual antigens ranged between 83.1% for EA-D and 92.3% for both EBNA-1 trunc and VCA p18 (Table 3). Sensitivity is very similar, and between 83.8% for EA-D and 96.3% for EBNA-1 trunc. Kappa statistics showed substantial agreements for EBNA-1 pep, ZEBRA and EA-D and almost perfect agreement for VCA p18 and EBNA-1 trunc (k = 0.64 to k = 0.89).

Only 2 of the reference assay seropositive sera did not react with any or only one of the EBV antigens; 86 (63.2%) were seropositive against all four EBV antigens (EBNA-1 trunc, VCA p18, ZEBRA, EA-D), 29 (21.3%) against 3 and 17 (12.5%) against 2. Nine reference assay negative sera showed antibody responses against at least one antigen; in 5 of these, antibodies against multiple EBV antigens were detected. Determining overall EBV seropositivity by a combination of the antigens showed optimum specificity (92.3%) and sensitivity (97.1%) by seropositivity against at least 2 out of 4 EBV proteins (Table 4). Using this algorithm, almost perfect agreement between both EBV assays (k = 0.90) was reached. Inclusion of either EBNA-1 trunc or peptide in the algorithm showed to be equally specific, but slightly more sensitive when including EBNA-1 trunc (Table 4).

CMV Monoplex Serology validation

CMV Monoplex Serology is based on four proteins: pp52, pp28, pp65 and pp150 N-terminus (pp150N) and was validated against two different reference assays, an anti-CMV IgG ELISA based on commercially available virion proteins (RP Va) and the Enzygnost anti-CMV IgG (RP Vb). Antibody detection against individual CMV antigens pp28, pp52 and pp150N showed high concordance with both reference assays detecting a maximum of 6 false-positives or false-negatives (Fig 1-Va/Vb). Specificity ranged between 95.4% and 99.0% and sensitivity ranged between 94.0% and 98.7% (Table 3). Kappa statistics indicated almost perfect agreement with both reference assays (kappa 0.90 to 0.96). Assay performance for CMV antigen pp65 was slightly poorer in comparison with the other antigens showing a higher overlap of measured antibody reactivities for reference assay positive and negative samples in both reference panels (Fig 1-Va/Vb). However, kappa statistics still indicated almost perfect agreement with both reference assays (Va, k = 0.83; Vb, k = 0.85) (Table 3).

Overall CMV serostatus was determined by seropositivity against at least two out of three CMV antigens (pp28, pp52, pp150N) as inclusion of pp65 did not additionally improve assay performance and is therefore dispensable. Specificity was 96.9% and 100.0% for the two reference panels, while sensitivity was 98.7% and 96.0%, respectively (Table 4).

Comparison of performance of HHV Monoplex and Multiplex Serology

The reference sera were analyzed both in monoplex (i.e. one pathogen) and in multiplex (i.e. multiple pathogens) format in order to compare assay performance. The performance of HHV Multiplex Serology based on sensitivity, specificity and kappa statistics was evaluated in comparison to the corresponding species-specific Monoplex Serology results (Fig 2). While some of the individual species-specific HHV assays showed slight differences in sensitivity and specificity between the monoplex and multiplex format, the overall statistical performance of the species-specific assays in Multiplex Serology did not change. Sensitivity and specificity for overall seropositivity for all HHV species exceeded 90% in both monoplex and multiplex format. A high concordance with the corresponding reference assays was maintained (k ≥ 0.85). A direct comparison of Monoplex versus Multiplex Serology performance was conducted using ICCs and showed good to excellent reliability (ICC: 0.82–0.99).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Comparison of statistical performance of HHV species-specific assays in monoplex (blue) and multiplex (orange) format.

Performance is shown for overall seropositivity (EBV, CMV) and single antigens for HSV-1 (gG), HSV-2 (mgGu) and VZV (gE). I-Vb indicate corresponding reference panels. Cohen’s kappa statistics are shown in percent to improve visualization. For direct comparison of Monoplex and Multiplex Serology performance on the corresponding reference panel, ICCs were calculated showing good to excellent reliability (ICC_HSV1: 0.91 (95% CI 0.88–0.93), ICC_HSV2: 0.93 (95% CI 0.89–0.95), ICC_VZV: 0.93 (95% CI 0.91–0.95), ICC_EBV: 0.87 (95% CI 0.84–0.90), ICC_{CMV RPVa}: 0.82 (95% CI 0.77–0.86), ICC_{CMV RPVb}: 0.99 (95%CI 0.99–1.00)). kappa: Cohen’s kappa. ICC: Intraclass correlation coefficient. CI: confidence interval.

https://doi.org/10.1371/journal.pone.0209379.g002